Keeping AI Online: Centre Exploring Nuclear Power And Dedicated Grid For Data Centres

With data centre technology rapidly advancing in India, the government is exploring the feasibility of providing consistent and dedicated power to large-scale data centre parks. This could include direct power lines from producers or even small, local nuclear plants to ensure uninterrupted supply.

One potential solution is to allow these data centres to establish independent power grids, enabling them to source energy directly from producers or from power-surplus states, regardless of their own state's energy status, as reported by The Economic Times.

The Ministry of Electronics and Information Technology (MeitY) has already engaged the Power Ministry to assess the feasibility of this proposal.

“We are studying all the options and the best possible way ahead,” the report said, quoting an official. “Whether an option will be feasible or not – cost-wise and infrastructure-wise – and ensuring minimal power loss will be a part of our study.”

Globally, nuclear power—known for producing minimal greenhouse gas emissions—is emerging as a vital energy source for tech giants. Over the past two months, companies like Google, Microsoft, OpenAI, and Amazon have turned to advanced nuclear reactors to support their growing AI workloads.

On 14 October, Google made headlines by signing the world's first corporate agreement to purchase nuclear energy from a consortium of small modular reactors (SMRs), which will be built by a startup called Kairos Power.

Just two days later, Amazon unveiled its ambitious plan, partnering with Energy Northwest to develop four advanced SMRs. The company also picked up a strategic stake in X-energy, a leading SMR and fuel developer.

In September, Microsoft solidified its commitment to nuclear energy by signing a 20-year power purchase agreement with Constellation, which will facilitate the restart of the Three Mile Island Unit 1 in Pennsylvania — a facility that has remained dormant since 2019.

The tech industry’s dash for nuclear energy is, in part, fuelled by the explosive growth of AI. For instance, a single ChatGPT query requires 2.9 watt-hours of electricity, in stark contrast to just 0.3 watt-hours for a typical Google search, according to the International Energy Agency (IEA).

In that difference lies the coming sea change in how the world at large will consume power — and how much that will cost.

India’s data landscape is also rapidly evolving, with projections indicating a significant role in the global data economy. Unlike real estate, data centres are measured by power consumption rather than land area.

According to ICRA, India’s data centre capacity is expected to double from 950 MW in FY24 to approximately 2,000 MW by FY27. To put this in perspective, the combined power needs of India’s data centres will soon resemble that of an entire city.

The demand for uninterrupted power will only grow as emerging technologies—AI, machine learning, cryptocurrency, and blockchain—require high-performance GPUs housed in data centres. These advanced processors, though compact, generate significant heat and necessitate advanced cooling systems, further driving power consumption.

“That, in turn, means that both the quality of power being supplied and the uninterrupted…supply will have to be maintained at the highest standards. With AI servers and now GPUs coming into the picture, the demand will be a lot higher,” the official said.